SIsiraben·Feb 22, 2021
> I need a hero, and by that, I mean that I need a usable methodology for building scaleable decentralized apps. Yes, you've heard that right. We don't need more "Ethereum killers" that can do 10x more tx/s than Ethereum. Those are useless.
Having worked on large Ethereum smart contracts (running a decentralized auction on chain) over two years, I fully agree with this. Ethereum needs better smart contract languages, better compilers and better dev tooling.
As our smart contract written in Vyper (for ease of auditing) grew larger and larger, eventually we failed to deploy on the testnet, and closer inspection reveal that the compiler was blowing up code size which I had to patch[0]. Down the line upstream further increased gas costs by emitting code that zeroed out every member of an array (note the EVM already has all memory set to 0). T
Read the rest
o support efficient insertion/deletion of bids we had to handroll a skip list on the blockchain(!), because there are no Vyper libraries.
On the language development side, things don't look too pretty either, I ended up having to write an EVM assembler from scratch[1], but deeper still are problems such as a lack of a separate call stack in EVM, so procedure calls always end up clunky and implementation-dependent. LLVM to EVM[2] is still experimental, so anyone who is writing a compiler will have to write codegen from scratch.
Perhaps eWASM will alleviate some of these concerns. But until then, the entire dev process for writing smart contracts is painful.
[0] https://github.com/vyperlang/vyper/pull/1488
[1] https://github.com/ActorForth/evm-assembler/blob/master/docs...
[2] https://github.com/etclabscore/evm_llvm